Imagine having a personal AI assistant that actually does things, not just answering questions. One you can message through Telegram, Discord, or Slack and ask to check your calendar, draft an email, browse the web, or run a script on your server.

And one where your conversations and data never leave infrastructure you control.

That's essentially what OpenClaw is.

OpenClaw, previously known as Moltbot and originally Clawdbot, is an open-source AI agent that runs on your own machine or server and connects to messaging apps you already use. Instead of being another chat interface, it acts more like an assistant capable of executing tasks, maintaining memory, and interacting with external tools.

This guide explains what OpenClaw is, how it works, and how to get it running on your own system.

What makes OpenClaw different from a regular chatbot?

Most AI tools are reactive. You open an app, ask a question, and get a response.

OpenClaw is designed to go further.

First, it can be proactive. You can configure reminders, scheduled briefings, and automated notifications.

Second, it can execute tasks. Depending on your configuration, OpenClaw can:

  • run shell commands
  • interact with APIs
  • browse websites
  • manage files
  • connect with external services

Third, it maintains persistent memory across conversations. Instead of resetting context each session, it builds context over time.

OpenClaw itself does not include a model. Instead it connects to whichever backend you prefer, including:

  • Anthropic Claude
  • OpenAI models
  • local models via Ollama

The platform also supports integrations with messaging apps including Telegram, Discord, Slack, WhatsApp, Signal, and Matrix.

How OpenClaw works

OpenClaw runs a gateway service that manages:

  • assistant memory
  • integrations
  • messaging connections
  • task execution
  • configuration and permissions

Messaging platforms connect to the gateway so conversations maintain context even if you switch platforms.

OpenClaw also supports extensions and tools that allow it to interact with external systems or run local commands.

What you'll need before getting started

To self-host OpenClaw you'll need a machine to run it on. This could be:

  • your own computer
  • a VPS
  • a dedicated server

For lightweight setups, modest resources are enough. Running local AI models or heavy sandboxing may require more RAM and CPU.

You'll also need:

  • an Anthropic or OpenAI API key (unless using a local model)
  • bot tokens for messaging platforms
  • basic terminal familiarity

If you're deploying on a remote server, our guide on SSH can help.

Setting up your server

If you're installing on Ubuntu or Debian, start by updating system packages:

sudo apt update && sudo apt upgrade -y

Installing OpenClaw

The easiest way to install OpenClaw is with the official one-line installer.

It works on macOS, Linux, and Windows and installs everything required automatically.

macOS / Linux

Run the following command:

# Works everywhere. Installs everything. You're welcome. 🦞
curl -fsSL https://openclaw.ai/install.sh | bash

Windows (PowerShell)

Run the installer in PowerShell:

# Works everywhere. Installs everything. You're welcome. 🦞
iwr -useb https://openclaw.ai/install.ps1 | iex

The installer downloads dependencies, installs Node.js if needed, and launches the OpenClaw onboarding process.

Once installation finishes, open the dashboard in your browser:

http://127.0.0.1:18789/

From there you can:

  • connect your AI provider
  • pair the Control UI
  • configure messaging integrations

Is Docker required?

Docker is optional.

Most users should start with the normal install flow above, especially when running OpenClaw on their own machine.

Docker is useful if you want:

  • an isolated gateway environment
  • easier deployment on servers
  • containerized workflows

Docker can also be used specifically for agent sandboxing, even if the OpenClaw gateway runs directly on the host.

Running OpenClaw with Docker

If you prefer a containerized deployment, OpenClaw provides a Docker setup script.

Make sure you have Docker and Docker Compose installed.

From the OpenClaw repository directory, run:

./docker-setup.sh

The script will:

  • build or pull the OpenClaw container image
  • run the onboarding wizard
  • start the gateway using Docker Compose
  • generate a gateway token

After setup completes, open:

http://127.0.0.1:18789/

Paste the generated token into the dashboard to finish pairing.

If you want to skip building the container locally and instead pull the official image, run:

export OPENCLAW_IMAGE="ghcr.io/openclaw/openclaw:latest"
./docker-setup.sh

This will download the prebuilt container image from GitHub Container Registry.

Docker sandboxing and agent isolation

OpenClaw can also use Docker for sandboxing.

When enabled, agent tools can run inside isolated containers rather than directly on your host system. This allows stricter control over filesystem access, networking, and permissions.

Sandboxing is particularly useful if your agents execute shell commands or modify files.

Connecting your first messaging platform

Telegram

Telegram is often the easiest integration to start with.

Search for @BotFather and run:

/newbot

Follow the prompts and copy the generated bot token.

Add the token in the OpenClaw dashboard under Telegram integration settings.

Discord

Visit the Discord Developer Portal:

https://discord.com/developers/applications

Create an application, add a bot, generate an invite link, and add it to your server.

Then enter the bot token in OpenClaw.

Slack

Create a Slack app via:

https://api.slack.com/apps

Assign required scopes, install it to your workspace, and enter the credentials in OpenClaw.

Configuring your AI model

Inside the OpenClaw dashboard, choose your AI provider.

If using hosted models, paste your API key and select a model.

If running locally, configure OpenClaw to connect to Ollama:

http://localhost:11434

Example model:

llama3

Conclusion

OpenClaw allows you to run a personal AI agent on infrastructure you control while connecting it to messaging platforms you already use.

For most users, the standard install flow is the simplest way to get started. Docker is available if you want containerization or sandboxing, but it's optional.

That flexibility makes OpenClaw easy to run whether you want it on your own machine, a home lab server, or a public VPS.

For larger always-on deployments, xTom provides enterprise-grade dedicated servers and colocation services. For smaller deployments, V.PS offers scalable NVMe-powered KVM VPS hosting suitable for running OpenClaw.

xTom also provides IP transit, shared hosting, and a full range of IT services to support your infrastructure needs.

Ready to discuss your infrastructure needs? Contact our team to explore the right solution for your projects.

Frequently asked questions about self-hosting OpenClaw

What is OpenClaw?

OpenClaw is a self-hosted AI agent that connects to messaging platforms like Telegram, Discord, and Slack and can execute tasks using external tools.

Do I need Docker to run OpenClaw?

No. Docker is optional. The standard install flow works on macOS, Windows, and Linux without containers.

What does the one-line installer do?

The installer installs Node.js, dependencies, and launches the OpenClaw onboarding process.

Can Docker be used only for sandboxing?

Yes. You can run the OpenClaw gateway directly on the host while using Docker only for sandboxed agent tools.

What messaging platforms does OpenClaw support?

OpenClaw supports Telegram, Discord, Slack, WhatsApp, Signal, Matrix, and other integrations.

What are Clawdbot, Moltbot, and OpenClaw?

They are the same project under different names. The project started as Clawdbot, was renamed Moltbot, and is now OpenClaw.